CN111814976B - Air conditioning system sensor fault error relearning method and system - Google Patents
Air conditioning system sensor fault error relearning method and system Download PDFInfo
- Publication number
- CN111814976B CN111814976B CN202010676072.3A CN202010676072A CN111814976B CN 111814976 B CN111814976 B CN 111814976B CN 202010676072 A CN202010676072 A CN 202010676072A CN 111814976 B CN111814976 B CN 111814976B
- Authority
- CN
- China
- Prior art keywords
- data
- neural network
- error
- prediction error
- relearning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 79
- 238000004378 air conditioning Methods 0.000 title claims abstract description 71
- 238000013528 artificial neural network Methods 0.000 claims abstract description 170
- 238000012549 training Methods 0.000 claims abstract description 46
- 230000008569 process Effects 0.000 claims description 9
- 230000007257 malfunction Effects 0.000 claims 3
- 230000000694 effects Effects 0.000 abstract description 4
- 238000012545 processing Methods 0.000 abstract description 4
- 238000003745 diagnosis Methods 0.000 description 25
- 238000012360 testing method Methods 0.000 description 5
- 238000001514 detection method Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F11/00—Control or safety arrangements
- F24F11/30—Control or safety arrangements for purposes related to the operation of the system, e.g. for safety or monitoring
- F24F11/32—Responding to malfunctions or emergencies
- F24F11/38—Failure diagnosis
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F11/00—Control or safety arrangements
- F24F11/50—Control or safety arrangements characterised by user interfaces or communication
- F24F11/56—Remote control
- F24F11/58—Remote control using Internet communication
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F11/00—Control or safety arrangements
- F24F11/62—Control or safety arrangements characterised by the type of control or by internal processing, e.g. using fuzzy logic, adaptive control or estimation of values
- F24F11/63—Electronic processing
- F24F11/64—Electronic processing using pre-stored data
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F11/00—Control or safety arrangements
- F24F11/89—Arrangement or mounting of control or safety devices
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F2110/00—Control inputs relating to air properties
- F24F2110/10—Temperature
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F24—HEATING; RANGES; VENTILATING
- F24F—AIR-CONDITIONING; AIR-HUMIDIFICATION; VENTILATION; USE OF AIR CURRENTS FOR SCREENING
- F24F2110/00—Control inputs relating to air properties
- F24F2110/20—Humidity
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Mechanical Engineering (AREA)
- Combustion & Propulsion (AREA)
- Chemical & Material Sciences (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Signal Processing (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Human Computer Interaction (AREA)
- Life Sciences & Earth Sciences (AREA)
- Fuzzy Systems (AREA)
- Air Conditioning Control Device (AREA)
Abstract
The invention discloses a method and a system for relearning fault errors of a sensor of an air conditioning system, which are characterized in that historical data of the sensor is used as training data to construct a basic neural network to obtain a prediction error; because the prediction error data is relatively smaller, the outlier data is deleted through standard deviation after the proportional expansion processing is carried out on the prediction error, so that the data distribution is stable; after relearning by using data with stable distribution, the obtained prediction structure is more accurate; finally, the relearned data is regressed to obtain real error data, so that the predicted data is more approximate to the actual data, the negative influence of residual error fault identification is reduced, the fault identification accuracy is higher, the error identification with smaller error degree is satisfied, and the method has wide applicability; on the basis of recognizing that errors exist, the invention predicts the faults by relearning the neural network, thereby achieving the effects of eliminating errors and improving accuracy, and the fault prediction accuracy is higher.
Description
Technical Field
The invention belongs to the technical field of air conditioner sensor fault diagnosis, and particularly relates to a method and a system for relearning fault errors of an air conditioner system sensor.
Background
Most of modern air conditioning systems rely on automatic control to meet user comfort and specified energy consumption requirements; when the air-conditioner control system fails, unnecessary energy waste is caused and the comfort requirement of a user is influenced; therefore, it is very important to detect and diagnose faults in air conditioning systems. The sensor is used as an important component of the air conditioner control system and directly determines the accuracy and precision of fault diagnosis of the air conditioner system; therefore, the fault diagnosis of the sensor is particularly important, especially, some small faults are difficult to detect by some current detection means, but faults with small fault degrees possibly cause large hidden troubles to the air conditioning system.
Currently, a BP neural network method, a wavelet analysis method, a fuzzy neural network method or the like is generally adopted for sensor fault diagnosis; the BP neural network method is a multi-layer feedforward neural network with forward signal transmission and backward error propagation; the BP neural network can learn and store a large number of mapping relations between input modes and output modes, and mathematical equations of the previous relations are not needed; the BP neural network comprises an input layer, an implicit layer and an output layer; the input data is subjected to standardization processing through the input layer and is applied with a corresponding weight threshold, the activation function is transferred to the hidden layer, the hidden layer is transferred to the output layer through the corresponding weight threshold and the activation function again, and the output layer outputs a corresponding neural network predicted value; if the predicted output data of the output layer does not reach the corresponding expected output data, performing an error counter-propagation stage; the output error is returned to each layer according to a certain mode, and the weight threshold value of each layer is modified; the prediction output data is output by continuous modification, namely, the learning process of the neural network, when the error is reduced to a certain acceptable degree or a certain learning times are cut off.
Currently, the BP neural network method is widely applied to fault diagnosis of an air conditioning system sensor, but still cannot have good effects on some small faults, because the current BP neural network has larger error value in prediction data, and if the error range of the error value is larger than the fault degree of the air conditioning system sensor, it is difficult to distinguish whether the error of the prediction result of the neural network data is caused by the self error or the fault of the air conditioning system sensor is caused by the error of the neural network data;
therefore, a method for reducing the prediction data error by a method on the premise of not changing the structure of the neural network is needed, and the method is more approximate to actual data; on the basis of the conventional BP neural network fault diagnosis method, the accuracy of detecting the sensor micro fault of the air conditioning system by using the neural network is improved.
Disclosure of Invention
Aiming at the technical problems in the prior art, the invention provides a method, a system and a storage medium for relearning the fault error of an air conditioner system sensor, which are used for solving the technical problems that the prior BP neural network has larger error value in predicted data, and further the fault diagnosis precision of the air conditioner sensor system is reduced.
In order to achieve the above purpose, the invention adopts the following technical scheme:
the invention provides a method for relearning fault errors of air conditioning system sensors, which comprises the following steps:
step 1, collecting sensor data of an air conditioning system in real time, and constructing a training sample; taking part of sensor data in the training sample as a group of sample data, and the rest of sensor data as a group of sample data;
step 2, constructing a basic neural network, and taking the b-group sample data into the basic neural network to train the basic neural network to obtain a trained basic neural network;
step 3, taking a part of the sample data in the group a as input data a 1 The remaining part of the data is taken as output data a 2 The method comprises the steps of carrying out a first treatment on the surface of the Will input data a 1 Is carried into a trained basic neural network to obtain predicted output data a Prediction ；
Step 4, utilizing the output data a 2 And predicting output data a Prediction Calculating to obtain a prediction error a Prediction error ；
Step 5, predicting the error a Prediction error Scaling up to a preset scaling up distribution interval (-n, n) to obtain a scaling up data set a';
step 6, eliminating the proportion expansion data aConstructing a sensor error data set a'; acquiring the sum prediction error a in the sensor error data set a' Prediction error Corresponding elements one by one and according to the prediction error a Prediction error Sequentially arranged corresponding to the sensor error data set a 'to obtain a new element set a' Predictive data ；
Step 7, constructing a relearning neural network and collecting a 'new element set' Predictive data The data in the training data are brought into a relearning neural network, and the relearning neural network is trained to obtain a trained relearning neural network;
step 8, taking the predicted output data in the step 3 as a Prediction As input data, the training re-learning neural network is brought into, and output data a of the re-learning neural network is obtained m The method comprises the steps of carrying out a first treatment on the surface of the Output data a for relearning neural network m Performing range inverse process to obtain prediction error regression data a Predictive regression ；
Step 9, using the predicted output data a Prediction Prediction error regression data a Predictive regression Calculating to obtain output data a Output of Output data a Output of The result of the air conditioning system sensor fault error relearning method is obtained; will output data a Output of Comparing with a preset threshold, outputting data a when outputting data Output of The data exceeding the threshold value part is the sensor fault data.
Further, in step 1, the sensor data of the air conditioning system includes fresh air temperature, fresh air humidity, air supply temperature, and air supply humidityDegree and return air temperature; in step 3, data a is input 1 The device comprises fresh air temperature, fresh air humidity, air supply temperature, air supply humidity and return air temperature; output data a 2 Including the return air temperature.
Further, in step 1, after the plurality of sensor data in the training sample are randomly disturbed, the plurality of sensor data are divided into a group of sample data and b group of sample data according to a ratio of 1:1.
Further, in step 2, the basic neural network adopts a BP neural network; in step 7, the relearning neural network adopts a BP neural network.
Further, in step 4, the prediction error a Prediction error The mathematical expression of (2) is:
prediction error a Prediction error Output data a = 2 -predicting output data a Prediction 。
Further, in step 5, the mathematical expression of the scaling-up data set a' is;
wherein a '(x) is the xth element in the scaled-up array a', a Prediction error (x) For prediction error a Prediction error The x element in (a) is n which is the upper limit value of the preset expansion distribution interval, a Prediction error max For prediction error a Prediction error Maximum value element, minimum value element a Prediction error min For prediction error a Prediction error Is the minimum element of (a).
Further, in step 8, the prediction error regression data a Predictive regression The mathematical expression of (2) is:
further, in step 9, the data a is outputted Output of The mathematical expression of (2) is:
a output of ＝a Prediction -a Predictive regression 。
The invention also provides a sensor fault error relearning system of the air conditioning system, which comprises a sensor data acquisition module, a basic neural network module and a relearning neural network module;
the sensor data acquisition module is used for acquiring sensor data of the air conditioning system in real time and constructing a training sample; taking part of sensor data in the training sample as a group of sample data, and the rest of sensor data as a group of sample data;
the basic neural network module is used for constructing a basic neural network, bringing the b groups of sample data into the basic neural network, and training the basic neural network to obtain a trained basic neural network; taking part of the sample data in the group a as input data a 1 The remaining part of the data is taken as output data a 2 The method comprises the steps of carrying out a first treatment on the surface of the Will input data a 1 Is carried into a trained basic neural network to obtain predicted output data a Prediction The method comprises the steps of carrying out a first treatment on the surface of the By means of the output data a 2 And predicting output data a Prediction Calculating to obtain a prediction error a Prediction error The method comprises the steps of carrying out a first treatment on the surface of the Will predict the error a Prediction error Scaling up to a preset scaling up distribution interval (-n, n) to obtain a scaling up data set a'; reject ratio-enlarged data aConstructing a sensor error data set a'; acquiring the sum prediction error a in the sensor error data set a' Prediction error Corresponding elements one by one and according to the prediction error a Prediction error Sequentially arranged corresponding to the sensor error data set a 'to obtain a new element set a' Predictive data ；
A relearning neural network module for constructing a relearning neural network and collecting new elements a' Predictive data The data in the training data are brought into a relearning neural network, and the relearning neural network is trained to obtain a trained relearning neural network; taking the predicted output data in the step 3 as a Prediction As an input numberAccording to the result, the trained relearning neural network is brought in to obtain the output data a of the relearning neural network m The method comprises the steps of carrying out a first treatment on the surface of the Output data a for relearning neural network m Performing range inverse process to obtain prediction error regression data a Predictive regression The method comprises the steps of carrying out a first treatment on the surface of the Using predicted output data a Prediction Prediction error regression data a Predictive regression Calculating to obtain output data a Output of Output data a Output of The result of the air conditioning system sensor fault error relearning method is obtained; will output data a Output of Comparing with a preset threshold, outputting data a when outputting data Output of The data exceeding the threshold value part is the sensor fault data.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a method and a system for relearning fault errors of a sensor of an air conditioning system, which are characterized in that a basic neural network is constructed by using historical data of the sensor as training data to obtain a prediction error; because the prediction error data is relatively small, the outlier data is deleted through standard deviation after the proportional expansion processing is carried out on the prediction error, so that the data distribution is stable; after relearning by using data with stable distribution, the obtained prediction structure is more accurate; and finally, the relearned data is regressed to obtain real error data, so that the predicted data is more approximate to the actual data, the negative influence generated by residual error fault identification is reduced, the fault identification accuracy is higher, the error identification with smaller error degree is satisfied, the wide applicability is realized, and the accuracy of fault diagnosis of the air conditioner sensor system is effectively improved.
According to the air conditioning system sensor fault error relearning method and system, on the basis of the existence of an admission error, the fault is predicted through the relearning neural network, so that the effects of eliminating the error and improving the accuracy are achieved, and the fault prediction accuracy is higher; the error with smaller error degree can be detected in the aspect of fault diagnosis, and the method can be applied to other neural networks, and has wide applicability; the air conditioning unit sensor fault error relearning method can also be applied to the diagnosis of air conditioning system faults, is not limited to the fault diagnosis of air conditioning unit sensors, and has universality.
Drawings
FIG. 1 is a graph showing the actual values of the percentage of the air conditioning system sensor return air humidity, and the percentage of the air conditioning system sensor return air humidity predicted by the conventional BP neural network and the relearning neural network according to the present invention in example 1;
FIG. 2 is a graph showing the relative error of the predicted return air humidity of the example 1 through the conventional BP neural network and the relearning neural network of the present invention;
FIG. 3 is a graph showing the relative error of the return air humidity predicted by the conventional BP neural network and the relearning neural network according to the present invention when the drift fault test with the drift degree of 0.05/unit is set from the sensor data of the 101 st group to the sensor data of the 300 rd group in example 2;
fig. 4 is a graph showing the relative error results of the return air humidity predicted by the conventional BP neural network and the relearning neural network of the present invention when the bias fault test was performed with a bias degree of 2% from the 101 st group to the 300 nd group sensor data in example 2.
Detailed Description
The present invention will be described in further detail with reference to the following embodiments.
The invention provides a method for relearning fault errors of air conditioning system sensors, which comprises the following steps:
step 1, collecting sensor data of an air conditioning system in real time, and constructing a training sample; taking part of sensor data in the training sample as a group of sample data, and the rest of sensor data as a group of sample data; the sensor data of the air conditioning system comprises fresh air temperature, fresh air humidity, air supply temperature, air supply humidity and return air temperature; after the sensor data in the training sample are randomly disturbed, the sensor data are divided into a group of data and b group of data according to the proportion of 1:1.
Step 2, constructing a basic neural network, wherein the basic neural network adopts a BP neural network; and the b groups of sample data are brought into the basic neural network, and basic neural network training is carried out to obtain the trained basic neural network.
Step 3, taking a part of the sample data in the group a as input data a 1 The remaining part of the data is taken as output data a 2 The method comprises the steps of carrying out a first treatment on the surface of the Will input data a 1 Is carried into a trained basic neural network to obtain predicted output data a Prediction The method comprises the steps of carrying out a first treatment on the surface of the Wherein the data a is input 1 The device comprises fresh air temperature, fresh air humidity, air supply temperature, air supply humidity and return air temperature; output data a 2 Including the return air temperature.
Step 4, utilizing the output data a 2 And predicting output data a Prediction Calculating to obtain a prediction error a Prediction error And find the prediction error a Prediction error Maximum value element a of (a) Prediction error max And a minimum value element a Prediction error min ；
Wherein the prediction error a Prediction error The mathematical expression of (2) is:
prediction error a Prediction error Output data a = 2 -predicting output data a Prediction 。
Step 5, predicting the error a Prediction error Scaling up to a preset scaling up distribution interval (-n, n) to obtain a scaling up data set a';
wherein the mathematical expression of the scaling-up data set a' is;
wherein a '(x) is the xth element in the scaled-up array a', a Prediction error (x) For prediction error a Prediction error The x-th element of the distribution interval is n which is the upper limit value of the preset expansion distribution interval.
Step 6, eliminating the proportion expansion data aConstructing a sensor error data set a'; acquisition of sensor error data set a'And prediction error a Prediction error Corresponding elements one by one and according to the prediction error a Prediction error The arrangement sequence corresponding to the sensor error data set a ' is arranged to obtain a new element set a ' ' Predictive data ；
Step 7, constructing a relearning neural network and collecting a 'new element set' Predictive data And (3) carrying out the training of the relearning neural network by taking the data in the relearning neural network to obtain the relearning neural network after the training.
Step 8, taking the predicted output data in the step 3 as a Prediction As input data, the training re-learning neural network is brought into, and output data a of the re-learning neural network is obtained m The method comprises the steps of carrying out a first treatment on the surface of the Output data a for relearning neural network m Performing range inverse process to obtain prediction error regression data a Predictive regression ；
Wherein the prediction error regression data a Predictive regression The mathematical expression of (2) is:
step 9, using the predicted output data a Prediction Prediction error regression data a Predictive regression Calculating to obtain output data a Output of Output data a Output of The result of the air conditioning system sensor fault error relearning method is obtained; will output data a Output of Comparing with a preset threshold, outputting data a when outputting data Output of The data exceeding the threshold value part is the sensor fault data; further acquiring fault information of a sensor of the air conditioning system;
wherein the data a is outputted Output of The mathematical expression of (2) is:
a output of ＝a Prediction -a Predictive regression 。
The invention also provides a sensor fault error relearning system of the air conditioning system, which comprises a sensor data acquisition module, a basic neural network module and a relearning neural network module;
the sensor data acquisition module is used for acquiring sensor data of the air conditioning system in real time and constructing a training sample; taking part of sensor data in the training sample as a group of sample data, and the rest of sensor data as a group of sample data;
the basic neural network module is used for constructing a basic neural network, bringing the b groups of sample data into the basic neural network, and training the basic neural network to obtain a trained basic neural network; taking part of the sample data in the group a as input data a 1 The remaining part of the data is taken as output data a 2 The method comprises the steps of carrying out a first treatment on the surface of the Will input data a 1 Is carried into a trained basic neural network to obtain predicted output data a Prediction The method comprises the steps of carrying out a first treatment on the surface of the By means of the output data a 2 And predicting output data a Prediction Calculating to obtain a prediction error a Prediction error And find the prediction error a Prediction error Maximum value element a of (a) Prediction error max And a minimum value element a Prediction error min The method comprises the steps of carrying out a first treatment on the surface of the Will predict the error a Prediction error Scaling up to a preset scaling up distribution interval (-n, n) to obtain a scaling up data set a'; reject ratio-enlarged data aConstructing a sensor error data set a'; acquiring the sum prediction error a in the sensor error data set a' Prediction error Corresponding elements one by one and according to the prediction error a Prediction error The arrangement sequence corresponding to the sensor error data set a ' is arranged to obtain a new element set a ' ' Predictive data ；
A relearning neural network module for constructing a relearning neural network and collecting new elements a' Predictive data The data in the training data are brought into a relearning neural network, and the relearning neural network is trained to obtain a trained relearning neural network; taking the predicted output data in the step 3 as a Prediction As input data, the training re-learning neural network is brought into, and output data a of the re-learning neural network is obtained m The method comprises the steps of carrying out a first treatment on the surface of the Output data a for relearning neural network m Performing range inverse process to obtain prediction error regression data a Predictive regression The method comprises the steps of carrying out a first treatment on the surface of the Using predicted output data a Prediction Prediction error regression data a Predictive regression Calculating to obtain output data a Output of Output data a Output of The result of the air conditioning system sensor fault error relearning method is obtained; will output data a Output of Comparing with a preset threshold, outputting data a when outputting data Output of The data exceeding the threshold value is sensor fault data, and further the air conditioning system sensor fault information is obtained.
According to the air conditioning system sensor fault error relearning method and system, the historical data of the sensor is used as training data, a basic neural network is constructed, and a prediction error is obtained; because the prediction error data is relatively small, the outlier data is deleted through standard deviation after the proportional expansion processing is carried out on the prediction error, so that the data distribution is stable; after relearning by using data with stable distribution, the obtained prediction structure is more accurate; and finally, carrying out regression on the relearned data to obtain real error data, so that the predicted data is more approximate to the actual data, the negative influence of residual error fault identification is reduced, the fault identification accuracy is higher, the error identification with smaller error degree is satisfied, and the method has wide applicability.
Example 1
In embodiment 1, 100 groups of sensor data in a central air conditioning system are acquired, and the sensor fault diagnosis is performed by using the air conditioning system sensor fault error relearning method, which specifically comprises the following steps:
step 1, acquiring 100 groups of sensor data of an air conditioning system in real time, wherein each group of sensor data comprises fresh air temperature, fresh air humidity, air supply temperature, air supply humidity and return air temperature of the air conditioning system at the same moment, and constructing a training sample; after the sensor data in the training samples are randomly disturbed, 50 groups of sensor data are used as a group of sample data according to the proportion of 1:1, and the rest 50 groups of data are used as b groups of sample data.
Step 2, constructing a basic neural network by adopting a BP neural network, wherein an input layer of the BP neural network comprises 5 input points, an output layer comprises 1 output point, and a hidden layer comprises 8 neurons; and the b groups of sample data are brought into the basic neural network, and basic neural network training is carried out to obtain the trained basic neural network.
Step 3, taking the fresh air temperature, fresh air humidity, air supply temperature, air supply humidity and return air temperature data of the air conditioning system at the same moment in the sample data of the group a as input data a 1 Taking the return air temperature data of the air conditioning system at the same moment in the sample data of the group a as output data a 2 The method comprises the steps of carrying out a first treatment on the surface of the Will input data a 1 Is carried into a trained basic neural network to obtain predicted output data a Prediction 。
Step 4, utilizing the output data a 2 And predicting output data a Prediction Calculating to obtain a prediction error a Prediction error And find the prediction error a Prediction error Maximum value element a of (a) Prediction error max And a minimum value element a Prediction error min ；
Wherein the prediction error a Prediction error The mathematical expression of (2) is:
prediction error a Prediction error Output data a = 2 -predicting output data a Prediction 。
Step 5, predicting the error a Prediction error Scaling up to a preset scaling up distribution interval (-n, n) to obtain a scaling up data set a';
wherein the mathematical expression of the scaling-up data set a' is;
wherein a '(x) is the xth element in the scaled-up array a', a Prediction error (x) For prediction error a Prediction error The x-th element of the distribution interval is n which is the upper limit value of the preset expansion distribution interval.
Step 6, eliminating the proportionIn the enlarged data aConstructing a sensor error data set a'; acquiring the sum prediction error a in the sensor error data set a' Prediction error Corresponding elements one by one and according to the prediction error a Prediction error Sequentially arranged corresponding to the sensor error data set a 'to obtain a new element set a' Predictive data ；
Step 7, constructing a relearning neural network and collecting a 'new element set' Predictive data And (3) carrying out the training of the relearning neural network by taking the data in the relearning neural network to obtain the relearning neural network after the training.
Step 8, taking the predicted output data in the step 3 as a Prediction As input data, the training re-learning neural network is brought into, and output data a of the re-learning neural network is obtained m The method comprises the steps of carrying out a first treatment on the surface of the Output data a for relearning neural network m Performing range inverse process to obtain prediction error regression data a Predictive regression ；
Wherein the prediction error regression data a Predictive regression The mathematical expression of (2) is:
step 9, using the predicted output data a Prediction Prediction error regression data a Predictive regression Calculating to obtain output data a Output of Output data a Output of The result of the air conditioning system sensor fault error relearning method is output data a Output of Comparing with a preset threshold, outputting data a when outputting data Output of The data exceeding the threshold value part is sensor fault data, and further sensor fault information of the air conditioning system is obtained;
wherein the data a is outputted Output of The mathematical expression of (2) is:
a output of ＝a Prediction -a Predictive regression 。
Under the same condition, 100 groups of sensor data in a certain central air conditioning system are acquired by utilizing the existing BP basic neural network pair, sensor fault diagnosis is carried out, the sensor fault data are acquired, and the sensor fault information is judged.
The actual value of the percentage of the return air humidity of the air conditioning system sensor, the data of the percentage of the return air humidity predicted by the existing BP neural network and the curve of the percentage of the return air humidity predicted by the relearning neural network according to the invention are shown in the attached figure 1, the percentage of the return air humidity obtained by the fault error relearning method according to the invention is closer to the actual value of the percentage of the return air humidity of the air conditioning system sensor, and the accuracy of the prediction result is higher.
The relative error curve of the return air humidity predicted by the conventional BP neural network and the relearning neural network is shown in the attached figure 2, and as can be seen from the attached figure 2, the relative error of the return air humidity predicted by the fault error relearning method is obviously smaller than that predicted by the conventional BP neural network, the average error is reduced from 0.0862 to 0.0690, and the average error is reduced by 19.95%.
Example 2
In this embodiment, 300 groups of sensor data in a central air conditioning system are obtained, and the sensor fault diagnosis is performed by using the air conditioning system sensor fault error relearning method, which specifically includes the following steps:
step 1, collecting 300 groups of sensor data of an air conditioning system in real time, wherein each group of sensor data comprises fresh air temperature, fresh air humidity, air supply temperature, air supply humidity and return air temperature of the air conditioning system at the same moment, and constructing a training sample; after the sensor data in the training samples are randomly disturbed, 150 groups of sensor data are used as a group of sample data according to the proportion of 1:1, and the rest 150 groups of data are used as b groups of sample data.
Step 2, constructing a basic neural network by adopting a BP neural network, wherein an input layer of the BP neural network comprises 5 input points, an output layer comprises 1 output point, and a hidden layer comprises 8 neurons; and the b groups of sample data are brought into the basic neural network, and basic neural network training is carried out to obtain the trained basic neural network.
Step 3, taking the fresh air temperature, fresh air humidity, air supply temperature, air supply humidity and return air temperature data of the air conditioning system at the same moment in the sample data of the group a as input data a 1 Taking the return air temperature data of the air conditioning system at the same moment in the sample data of the group a as output data a 2 The method comprises the steps of carrying out a first treatment on the surface of the Will input data a 1 Is carried into a trained basic neural network to obtain predicted output data a Prediction 。
Step 4, utilizing the output data a 2 And predicting output data a Prediction Calculating to obtain a prediction error a Prediction error And find the prediction error a Prediction error Maximum value element a of (a) Prediction error max And a minimum value element a Prediction error min ；
Wherein the prediction error a Prediction error The mathematical expression of (2) is:
prediction error a Prediction error Output data a = 2 -predicting output data a Prediction 。
Step 5, predicting the error a Prediction error Scaling up to a preset scaling up distribution interval (-n, n) to obtain a scaling up data set a';
wherein the mathematical expression of the scaling-up data set a' is;
wherein a '(x) is the xth element in the scaled-up array a', a Prediction error (x) For prediction error a Prediction error The x-th element of the distribution interval is n which is the upper limit value of the preset expansion distribution interval.
Step 6, eliminating the proportion expansion data aConstructing a sensor error data set a'; acquiring the sum prediction error a in the sensor error data set a' Prediction error Corresponding elements one by one and according to the prediction error a Prediction error Sequentially arranged corresponding to the sensor error data set a 'to obtain a new element set a' Predictive data ；
Step 7, constructing a relearning neural network and collecting a 'new element set' Predictive data And (3) carrying out the training of the relearning neural network by taking the data in the relearning neural network to obtain the relearning neural network after the training.
Step 8, taking the predicted output data in the step 3 as a Prediction As input data, the training re-learning neural network is brought into, and output data a of the re-learning neural network is obtained m The method comprises the steps of carrying out a first treatment on the surface of the Output data a for relearning neural network m Performing range inverse process to obtain prediction error regression data a Predictive regression ；
Wherein the prediction error regression data a Predictive regression The mathematical expression of (2) is:
step 9, using the predicted output data a Prediction Prediction error regression data a Predictive regression Calculating to obtain output data a Output of Output data a Output of The result of the air conditioning system sensor fault error relearning method is obtained; will output data a Output of Comparing with a preset threshold, outputting data a when outputting data Output of The data exceeding the threshold value part is the sensor fault data; further acquiring fault information of a sensor of the air conditioning system;
wherein the data a is outputted Output of The mathematical expression of (2) is:
a output of ＝a Prediction -a Predictive regression 。
Under the same condition, 300 groups of sensor data in a certain central air conditioning system are acquired by utilizing the existing BP neural network pair, sensor fault diagnosis is carried out, the sensor fault data are acquired, and the sensor fault information is judged.
FIG. 3 is a graph showing the relative error results of the return air humidity predicted by the conventional BP neural network and the relearning neural network according to the invention when the drift fault test with the drift degree of 0.05/unit is set according to the sensor data from the 101 st group to the 300 rd group; as can be seen from fig. 3, under the fault test condition that the drift degree is 0.05/unit, the diagnosis success rate of the existing BP neural network is distributed between 29% and 47%, the diagnosis success rate of the relearning neural network is distributed between 31% and 67%, the diagnosis success rate is generally improved by 4% to 22%, the average detection accuracy rate is improved by 48.47% from 37.91%, and the improvement is 10.44%.
FIG. 4 is a graph showing the relative error results of the return air humidity predicted by the conventional BP neural network and the relearning neural network according to the invention when the bias fault test is performed by setting the bias degree to be 2% according to the sensor data from the 101 st group to the 300 rd group; as can be seen from fig. 4, when a drift fault with a deviation degree of 2% is set, the diagnosis rate of the existing BP neural network is low; the fault triggering threshold of 0.4 is used as a demarcation limit, the diagnosis success rate of the conventional BP neural network is distributed between 35% and 49.5%, the diagnosis success rate of the learning neural network is distributed between 36% and 58%, the diagnosis success rate is generally improved by 2% to 11.5%, the average detection accuracy rate is improved from 41.5% to 49.1%, and the improvement is 7.6%.
According to the air conditioning system sensor fault error relearning method and system, on the basis of the existence of an admission error, the fault is predicted through the relearning neural network, so that the effects of eliminating the error and improving the accuracy are achieved, and the fault prediction accuracy is higher; the error with smaller error degree can be detected in the aspect of fault diagnosis, and the method can be applied to other neural networks, and has wide applicability; the air conditioning unit sensor fault error relearning method can also be applied to the diagnosis of air conditioning system faults, is not limited to the fault diagnosis of air conditioning unit sensors, and has universality.
The foregoing merely illustrates the preferred embodiments of the invention and any structural modifications, improvements, or adaptations of the invention that may be made by any person without departing from the principles of the invention are intended to be within the scope of the invention.
Claims (8)
1. The air conditioning system sensor fault error relearning method is characterized by comprising the following steps of:
step 1, collecting sensor data of an air conditioning system in real time, and constructing a training sample; taking part of sensor data in the training sample as a group of sample data, and the rest of sensor data as a group of sample data;
step 2, constructing a basic neural network, and taking the b-group sample data into the basic neural network to train the basic neural network to obtain a trained basic neural network;
step 3, taking a part of the sample data in the group a as input data a 1 The remaining part of the data is taken as output data a 2 The method comprises the steps of carrying out a first treatment on the surface of the Will input data a 1 Is carried into a trained basic neural network to obtain predicted output data a Prediction ；
Step 4, utilizing the output data a 2 And predicting output data a Prediction Calculating to obtain a prediction error a Prediction error ；
Step 5, predicting the error a Prediction error Scaling up to a preset scaling up distribution interval (-n, n) to obtain a scaling up data set a'; wherein the mathematical expression of the scaling-up data set a' is;
wherein a '(x) is the xth element in the scaled-up array a', a Prediction error (x) For prediction error a Prediction error The x element in (a) is n which is the upper limit value of the preset expansion distribution interval, a Prediction error max For prediction error a Prediction error Maximum of (3)Value element, minimum value element a Prediction error min For prediction error a Prediction error Minimum value element in (2)
Step 6, eliminating the proportion expansion data aConstructing a sensor error data set a'; acquiring the sum prediction error a in the sensor error data set a' Prediction error Corresponding elements one by one and according to the prediction error a Prediction error Sequentially arranged corresponding to the sensor error data set a 'to obtain a new element set a' Predictive data ；
Step 7, constructing a relearning neural network and collecting a 'new element set' Predictive data The data in the training data are brought into a relearning neural network, and the relearning neural network is trained to obtain a trained relearning neural network;
step 8, taking the predicted output data in the step 3 as a Prediction As input data, the training re-learning neural network is brought into, and output data a of the re-learning neural network is obtained m The method comprises the steps of carrying out a first treatment on the surface of the Output data a for relearning neural network m Performing range inverse process to obtain prediction error regression data a Predictive regression ；
Step 9, using the predicted output data a Prediction Prediction error regression data a Predictive regression Calculating to obtain output data a Output of Output data a Output of The result of the air conditioning system sensor fault error relearning method is obtained; will output data a Output of Comparing with a preset threshold, outputting data a when outputting data Output of The data exceeding the threshold value part is the sensor fault data.
2. The method for relearning sensor fault errors in an air conditioning system according to claim 1, wherein in step 1, the sensor data of the air conditioning system includes fresh air temperature, fresh air humidity, supply air temperature, supply air humidity and return air temperature; in step 3, data a is input 1 The device comprises fresh air temperature, fresh air humidity, air supply temperature, air supply humidity and return air temperature; output data a 2 Including the return air temperature.
3. The air conditioning system sensor fault error relearning method according to claim 1, wherein in step 1, after a plurality of sensor data in a training sample are randomly disturbed, the plurality of sensor data are divided into a group of sample data and b group of sample data according to a ratio of 1:1.
4. The air conditioning system sensor fault error relearning method of claim 1, wherein in step 2, the base neural network adopts a BP neural network; in step 7, the relearning neural network adopts a BP neural network.
5. The air conditioning system sensor malfunction error relearning method as set forth in claim 1, wherein in step 4, the prediction error a is Prediction error The mathematical expression of (2) is:
prediction error a Prediction error Output data a = 2 -predicting output data a Prediction 。
6. The air conditioning system sensor malfunction error relearning method as set forth in claim 1, wherein in step 8, the prediction error regression data a Predictive regression The mathematical expression of (2) is:
7. the air conditioning system sensor malfunction error relearning method as set forth in claim 1, wherein in step 9, the output data a Output of The mathematical expression of (2) is:
a output of ＝a Prediction -a Predictive regression 。
8. The air conditioning system sensor fault error relearning system is characterized by comprising a sensor data acquisition module, a basic neural network module and a relearning neural network module;
the sensor data acquisition module is used for acquiring sensor data of the air conditioning system in real time and constructing a training sample; taking part of sensor data in the training sample as a group of sample data, and the rest of sensor data as a group of sample data;
the basic neural network module is used for constructing a basic neural network, bringing the b groups of sample data into the basic neural network, and training the basic neural network to obtain a trained basic neural network; taking part of the sample data in the group a as input data a 1 The remaining part of the data is taken as output data a 2 The method comprises the steps of carrying out a first treatment on the surface of the Will input data a 1 Is carried into a trained basic neural network to obtain predicted output data a Prediction The method comprises the steps of carrying out a first treatment on the surface of the By means of the output data a 2 And predicting output data a Prediction Calculating to obtain a prediction error a Prediction error The method comprises the steps of carrying out a first treatment on the surface of the Will predict the error a Prediction error Scaling up to a preset scaling up distribution interval (-n, n) to obtain a scaling up data set a'; reject ratio-enlarged data aConstructing a sensor error data set a'; acquiring the sum prediction error a in the sensor error data set a' Prediction error Corresponding elements one by one and according to the prediction error a Prediction error Sequentially arranged corresponding to the sensor error data set a 'to obtain a new element set a' Predictive data ；
Wherein the mathematical expression of the scaling-up data set a' is;
wherein a' (x) isThe x-th element, a, in the scaling-up array a Prediction error (x) For prediction error a Prediction error The x element in (a) is n which is the upper limit value of the preset expansion distribution interval, a Prediction error max For prediction error a Prediction error Maximum value element, minimum value element a Prediction error min For prediction error a Prediction error The minimum element of (2);
a relearning neural network module for constructing a relearning neural network and collecting new elements a' Predictive data The data in the training data are brought into a relearning neural network, and the relearning neural network is trained to obtain a trained relearning neural network; taking the predicted output data as a Prediction As input data, the training re-learning neural network is brought into, and output data a of the re-learning neural network is obtained m The method comprises the steps of carrying out a first treatment on the surface of the Output data a for relearning neural network m Performing range inverse process to obtain prediction error regression data a Predictive regression The method comprises the steps of carrying out a first treatment on the surface of the Using predicted output data a Prediction Prediction error regression data a Predictive regression Calculating to obtain output data a Output of Output data a Output of The result of the air conditioning system sensor fault error relearning method is obtained; will output data a Output of Comparing with a preset threshold, outputting data a when outputting data Output of The data exceeding the threshold value part is the sensor fault data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010676072.3A CN111814976B (en) | 2020-07-14 | 2020-07-14 | Air conditioning system sensor fault error relearning method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010676072.3A CN111814976B (en) | 2020-07-14 | 2020-07-14 | Air conditioning system sensor fault error relearning method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111814976A CN111814976A (en) | 2020-10-23 |
| CN111814976B true CN111814976B (en) | 2024-04-09 |
Family
ID=72864756
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010676072.3A Active CN111814976B (en) | 2020-07-14 | 2020-07-14 | Air conditioning system sensor fault error relearning method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111814976B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022126397A1 (en) * | 2020-12-15 | 2022-06-23 | 深圳市大疆创新科技有限公司 | Data fusion method and device for sensor, and storage medium |
| CN113188235A (en) * | 2021-05-28 | 2021-07-30 | 西安建筑科技大学 | Multi-source fault diagnosis method, system and equipment for sensor in air conditioning system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015158198A1 (en) * | 2014-04-17 | 2015-10-22 | 北京泰乐德信息技术有限公司 | Fault recognition method and system based on neural network self-learning |
| WO2015172560A1 (en) * | 2014-05-16 | 2015-11-19 | 华南理工大学 | Central air conditioner cooling load prediction method based on bp neural network |
| CN108507117A (en) * | 2017-10-13 | 2018-09-07 | 上海智容睿盛智能科技有限公司 | A kind of Air-conditioning system sensor method for diagnosing faults based on wavelet neural network |
| CN108592352A (en) * | 2018-05-09 | 2018-09-28 | 国家计算机网络与信息安全管理中心 | The prediction technique of water cooling air conditioning system operating status health degree, apparatus and system |
| CN109582657A (en) * | 2018-11-28 | 2019-04-05 | 国家电网有限公司 | Capacitor faults prediction technique based on machine learning and big data analysis |
| CN109657789A (en) * | 2018-12-06 | 2019-04-19 | 重庆大学 | Gear case of blower failure trend prediction method based on wavelet neural network |
| CN110245745A (en) * | 2019-05-30 | 2019-09-17 | 华中科技大学 | Equipment remaining life prediction technique based on integrated bi-directional Recognition with Recurrent Neural Network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11379284B2 (en) * | 2018-03-13 | 2022-07-05 | Nec Corporation | Topology-inspired neural network autoencoding for electronic system fault detection |
-
2020
- 2020-07-14 CN CN202010676072.3A patent/CN111814976B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015158198A1 (en) * | 2014-04-17 | 2015-10-22 | 北京泰乐德信息技术有限公司 | Fault recognition method and system based on neural network self-learning |
| WO2015172560A1 (en) * | 2014-05-16 | 2015-11-19 | 华南理工大学 | Central air conditioner cooling load prediction method based on bp neural network |
| CN108507117A (en) * | 2017-10-13 | 2018-09-07 | 上海智容睿盛智能科技有限公司 | A kind of Air-conditioning system sensor method for diagnosing faults based on wavelet neural network |
| CN108592352A (en) * | 2018-05-09 | 2018-09-28 | 国家计算机网络与信息安全管理中心 | The prediction technique of water cooling air conditioning system operating status health degree, apparatus and system |
| CN109582657A (en) * | 2018-11-28 | 2019-04-05 | 国家电网有限公司 | Capacitor faults prediction technique based on machine learning and big data analysis |
| CN109657789A (en) * | 2018-12-06 | 2019-04-19 | 重庆大学 | Gear case of blower failure trend prediction method based on wavelet neural network |
| CN110245745A (en) * | 2019-05-30 | 2019-09-17 | 华中科技大学 | Equipment remaining life prediction technique based on integrated bi-directional Recognition with Recurrent Neural Network |
Non-Patent Citations (4)
| Title |
|---|
| Sensor fault detection and its efficiency analysis in air handling unit using the combined neural networks;Du zhiming etal.;Energy and buildings;20141130;全文 * |
| 基于BP神经网络优化的离心式冷水机组故障诊断研究;梁晴晴等;制冷与空调;20160131;第16卷(第1期);全文 * |
| 基于GA-BP神经网络建筑负荷预测模型;雷磊;闫秀英;;现代物业(中旬刊);20191215(第12期);全文 * |
| 基于遗传算法-误差反传神经网络的变风量空调系统空调箱故障诊断;张善兴;何金凝;杜志敏;何军;晋欣桥;;制冷技术;20191015(第05期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111814976A (en) | 2020-10-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US6609036B1 (en) | Surveillance system and method having parameter estimation and operating mode partitioning | |
| CN111814976B (en) | Air conditioning system sensor fault error relearning method and system | |
| EP3250965B1 (en) | Methods and systems for detecting, classifying and/or mitigating sensor error | |
| US20140012791A1 (en) | Systems and methods for sensor error detection and compensation | |
| CN111723925B (en) | Fault diagnosis method, device, equipment and medium for on-road intelligent train air conditioning unit | |
| Wang et al. | Fault detection and diagnosis for multiple faults of VAV terminals using self-adaptive model and layered random forest | |
| Kanal | Perceptron | |
| CN111103325A (en) | Electronic nose signal drift compensation method based on integrated neural network learning | |
| CN114740361A (en) | Fuel cell voltage prediction method based on long-short term memory neural network model | |
| CN113837364A (en) | Sewage treatment soft measurement method and system based on residual error network and attention mechanism | |
| CN116952654B (en) | Environment monitoring and early warning system for administrative supervision | |
| US6292738B1 (en) | Method for adaptive detection of engine misfire | |
| CN112541526A (en) | Electronic nose gas concentration prediction method based on PSO-ABC-ELM | |
| CN112269367B (en) | Fault detection method and system based on EEMD-combined neural network air processing system | |
| CN112734186B (en) | Method, system and storage medium for real-time assessment of air microbial pollution of train carriage | |
| CN112819087B (en) | Method for detecting abnormality of BOD sensor of outlet water based on modularized neural network | |
| CN116907772A (en) | Self-diagnosis and fault source identification method and system of bridge structure monitoring sensor | |
| KR100973527B1 (en) | Fault diagnosis apparatus for gas sensor and method at the same | |
| Dong et al. | Neural networks for engine fault diagnostics | |
| Bilski | Analysis of the classifier fusion efficiency in the diagnostics of the accelerometer | |
| Pelliccioni et al. | Use of neural net models to forecast atmospheric pollution | |
| Jha et al. | On detection of out of distribution inputs in deep neural networks | |
| CN116467645B (en) | Pollution source collection monitoring system | |
| CN117470274A (en) | On-line fault detection and repair method and system for multi-navigation sensor system | |
| CN117950390A (en) | Method, device, equipment and storage medium for detecting abnormal dormancy of vehicle |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |